Utilizing Artificial Intelligence in an HR Setting

Conversations about artificial intelligence (AI) have been everywhere recently. Congress has held hearings about it, news outlets have written articles about it and content providers have sampled AI’s abilities by having it draft something for them. But how does AI impact companies and human resources?

According to the Pew Research Center, 62% of Americans believe AI will have a major impact on workers, but only 28% believe it will impact them directly. Unfortunately, AI is already impacting employees — 4,000 jobs were lost in May 2023 due to AI, the first time it has been listed as a reason for a layoff.

While most of the recent conversation involves AI-generated content, other AI formats have been used in the workplace for a while.

AI in the Hiring Process

AI tools used in the hiring process have been praised for saving hiring managers valuable time and creating a diverse pool of applicants by removing bias from the initial review process. However, concerns have been raised about unintentional bias being built into these tools.

Resume review tools can use predictive analysis to determine what candidate profile would be the best fit for an open position, and then compare received electronic resumes to find the “best available” candidates. However, if a candidate uses certain words or phrases which may not fit the AI tool’s expectations, the candidate will receive a lower evaluation for no real reason.

More concerning are tools that analyze an applicant’s personality, knowledge and communication skills through recorded responses to interview questions and facial expressions. These tools assess a candidate’s fit for a job by matching them to a profile of the company’s “ideal employee” using appearance, communication skills, speech patterns, body language, personality, etc. However, some of these tools have been found to be biased — eliminating people of certain genders, races, ethnicities and disabilities by giving lower scores for factors (such as facial structure, accents, hair style, or wearing glasses or head scarves) that do not match the “ideal” parameters in the programming.

Regulations on the use of these tools are already in place. On April 25, 2023, four federal agencies — US Equal Employment Opportunity Commission (EEOC), Department of Justice (DOJ), Consumer Financial Protection Bureau (CFPB) and Federal Trade Commission (FTC) — issued a joint statement addressing concerns about AI and its potential impacts. The statement covered several topics, including defining AI, acknowledging its potential positive uses and negative impacts, highlighting potential areas for discrimination and affirming each agency’s commitment “to monitor the development and usage of automated systems and promote responsible innovation” as well as “pledge to vigorously use our collective authorities to protect individuals’ rights regardless of whether legal violations occur through traditional means or advanced technologies.”

At the state level, Illinois, Maryland and New York City have already passed laws regulating the use of “automated employment decisions tools” in the hiring process, with many other states and cities considering similar laws.

AI-Generated Content

Most of the latest news is around chatbots — conversational “chat” software — and the AI-generated content they produce. Chatbots such as OpenAI’s ChatGPT, Microsoft’s Bing and Google’s Bard are now available to the public by simply downloading the software or phone app and setting up an account.

In the workplace, chatbots can be used to research topics and generate content such as policies, procedures, emails, letters and disciplinary action. On the positive side, AI used for HR purposes can help effectively address legalities, uncomfortable topics and messages for general audiences. However, AI has also been shown to generate content which lacks empathy, is non-specific, disregards the privacy of others, does not offer face-to-face interactions or contradicts itself. Asking the same question in different ways could give different results, which could complicate or confuse the issue more.

Beyond these concerns are the inherent limitations of chatbots as they are built on a Large Language Model, which relies on many available data sources. However, the end results are only as accurate and valid as the data it references. For example, Wikipedia is an often-used resource but, since it relies on user-generated content, it has been proven to be only 80% accurate. In some cases, chatbots have also created their own inaccurate reference material used for developing and validating answers.

To build its database, chatbots retain all entered information for future reference by any user. Since users must input specific information to get the best results, they may need to enter sensitive, confidential information or even trade secrets — all of which is added to the chatbot’s database. Depending on the information entered and/or the parameters entered by a future user, companies may find their confidential data available to anyone asking the right questions.

Suggested Action To Take Before Using AI

As tools develop and improve, AI will find a place in most workplaces. To determine how AI will be allowed in your workplace, consider taking the following actions:

Do your research: Understand what defines AI, as well as the advantages and drawbacks to each tool. In addition to the information linked in this article, consider reading other resources to educate yourself as much as possible.

Some useful articles we found were the Chatbot Primer (5-part series) and Prompts for More Effective Chatbot Results from The New York Times, and also Conductor’s article about using a chatbot.

Research your tools: Learn how AI is incorporated into tools you are already using or considering using. If you choose to use AI tools, be sure to understand their validity and limitations. For example, if you are going to use virtual analysis of recorded interviews, understand the science behind it, including if the tool has been properly tested to remove implicit biases.

Establish policies and procedures on AI use: Draft a policy to outline when and how AI can and cannot be used. Include clear statements prohibiting discrimination and revealing confidential information. While the policy can be general to cover any AI, develop exact procedures and expectations as you initiate AI tools.

Train employees and managers: As you expand the use of AI tools in your company, train your employees and managers on proper and legal usage and expectations — such as reviewing and fact-checking all content before releasing it or personalizing a letter to an employee or customer.

Affinity HR Group will continue to monitor this emerging technology and the related regulations around its development and use.

By Paige McAllister, SPHR, SHRM-SCP